An Analytical Hierarchy Process-Integrated GIS Approach for Mapping Landslide Susceptibility in Kullu District, India

Abstract

Aims: To integrate the weighted thematic layers within a GIS environment to generate and validate a spatial landslide susceptibility map. Study Design: A GIS-based analytical study integrating the Analytical Hierarchy Process (AHP) was undertaken for spatial assessment and mapping of landslides. The methodology involved the identification of landslide conditioning factors, AHP-based weighting of factors and sub-factors, spatial integration of weighted thematic layers, and classification and validation of the resulting susceptibility map. Place and Duration of Study: The study was conducted in Kullu district, Himachal Pradesh, India, covering an area of approximately 5,503 km² and extending between 31°20′–32°25′ N and 76°56′–77°52′ E. The study was conducted from January to August 2026. Methodology: Twelve conditioning factors were selected based on a systematic literature review, expert opinion, field investigations, and historical landslide inventories: slope gradient, Topographic Wetness Index (TWI), drainage density, relative relief, Stream Power Index (SPI), geomorphology, distance from roads, curvature, distance from lineaments, elevation, slope aspect and Normalized Difference Vegetation Index (NDVI). Pairwise comparisons of the selected factors and their sub-factors were performed using Saaty's 1–9 AHP scale. The weighted thematic layers were subsequently integrated into a GIS environment to generate the final landslide susceptibility map. A total of 37 landslide locations were identified from the Geological Survey of India (GSI) inventory, of which seven were used for validation alongside 100 randomly selected non-landslide points located outside 200 m buffers around identified landslides. Results: The 12 causative factors showed substantial spatial variability across Kullu district, reflecting diverse topographic, hydrological, structural, geomorphological and vegetation conditions. AHP assigned the highest weights to slope (0.2559), TWI (0.1932) and drainage density (0.1459), together contributing 59.5% of the total weight, while relative relief and SPI were moderately important. The consistency ratio (0.0616) indicated acceptable consistency. Very low, low, medium, high and very high susceptibility zones covered 22.74%, 31.18%, 24.76%, 13.57% and 7.74%, respectively. ROC analysis demonstrated excellent model performance (AUC = 0.955), with 80.49% sensitivity and 89.94% specificity. The findings highlight the importance of terrain steepness and hydrological conditions in determining landslide susceptibility and provide a useful spatial screening tool for prioritising field investigations, slope stabilisation, drainage management, and environmentally informed land-use planning in the Himalayan landscape. Because the validation included only seven landslide locations, the reported AUC and associated performance metrics should be interpreted cautiously.